Fabrication of PES/PVP Water Filtration Membranes Using Cyrene®, a Safer Bio-Based Polar Aprotic Solvent

A more sustainable dialysis and water filtration membrane has been developed, by using the new, safer, bio-based solvent Cyrene® in place of N-methyl pyrrolidinone (NMP). The effects of solvent choice, solvent evaporation time, the temperature of casting gel, and coagulation bath together with the additive concentration on porosity and pore size distribution were studied. The results, combined with infrared spectra, SEM images, porosity results, water contact angle (WCA), and water permeation, confirm that Cyrene® is better media to produce polyethersulfone (PES) membranes. New methods, Mercury Intrusion Porosimetry (MIP) and NMR-based pore structure model, were applied to estimate the porosity and pore size distribution of the new membranes produced for the first time with Cyrene® and PVP as additive. Hansen Solubility Parameters in Practice (HSPiP) was used to predict polymer-solvent interactions. The use of Cyrene® resulted in reduced polyvinylpyrrolidone (PVP) loading than required when using NMP and gave materials with larger pores and overall porosity. Two different conditions of casting gel were applied in this study: a hot (70°C) and cold gel (17°C) were cast to obtain membranes with different morphologies and water filtration behaviours.


Introduction
The growing use of filtration membranes is the result of increasing attention paid to environmental problems linked to the availability of and growing demand for clean water [1][2][3]. Recent work has highlighted the need for new solvents in the membrane manufacturing process, in order to minimise the problems of toxic solvent release to the environment, with Polarclean5, DMSO Evol5, and ionic liquids being successfully applied in this process [4][5][6][7][8]. Porous polymeric membranes for water and wastewater treatment have traditionally been produced from cellulose acetate [9], polysulfone [10], polyvinylidene fluoride (PVDF) [11,12], polyacrylonitrile (PAN) [13], and polyvinyl alcohol (PVA) [14]. Polyethersulfone (PES) is used in separation due to its excellent performance and high thermal, chemical, hydraulic, and mechanical stability. The polymer generates membranes ranging from nanofiltration (NF) and Reverse Osmosis (RO) used for desalinisation processes, to ultrafiltration (UF) used for food, metal, and textile industry and microfiltration (MF) for purification of beverages, and separation of oil and water emulsions. Fouling of the membrane is as a result of the hydrophobicity of PES, making separation unpredictable and shortening its lifetime due to a higher energy demand to push the water through the pores [15,16]. To reduce the fouling effect, the PES membranes are modified via bulk modification, surface modification, and blending (a type of surface modification), with the latter being the most widely used method [17].
Filtration membranes have previously been fabricated using a casting solution of polyethersulfone (PES) in the repro-toxic solvents NMP [18][19][20], DMF [21], DMAc [22], or mixture of solvents [23,24] along with a hydrophilic homopolymer, such as polyvinylpyrrolidone (PVP) [21,24] or poly(ethylene glycol) (PEG) [25,26], resulting in membranes less prone to fouling [25,27]. DMSO has been utilised 2 Advances in Polymer Technology as the casting solvent [2,28,29], which is not in itself toxic, but it acts as a permeability enhancer and transfer compounds through the skin barrier. It is used in medicine as a skin adsorption enhancer for different drugs [30,31]. Recently NMP was added to REACH's restricted substances list [32], due to the reproductive effects of this solvent. NMP and dichloromethane (DCM) were linked to foetal and adult deaths and their use has been banned or limited from paint stripping products by the big retailers. [33] DMF and DMAc are also on the Substances of Very High Concern (SVHC) list as toxic for reproduction (may damage the unborn child). The list of available and safe polar aprotic solvents is thus getting smaller.
The pore forming and structure controlling agents PVP and PEG are incorporated into PES matrix, similar to other polymer additives reported in the literature [20,34,35], so as to increase the hydrophilicity, diffusion properties (due to pore size distribution on the membrane surface) antifouling, and hemocompatibility properties of PES membranes [36][37][38]. These asymmetric membranes can be modified by adjusting any of the following factors: the composition of the casting solution (additives; solvent), the solvent evaporation temperature/time, and the method utilised. In this study we aim to substitute the traditional solvent used (NMP) with a greener, safer alternative. Cyrene5 (dihydrolevoglucosenone) is a new commercially available, nontoxic solvent obtained from renewable waste and nonfood cellulosic source feedstock (e.g., straw, bagasse, and sawdust) (see Scheme 1). It is quickly proving to be a viable alternative polar aprotic solvent to both NMP and DMF [39][40][41][42][43].
Cyrene5 is a polar aprotic solvent similar to NMP, but renewable, biodegradable, and nontoxic [39]. Cyrene5 has been only given a hazard warning for being an eye irritant (E319) receiving an ECHA level 7 certification and has recently received permission to be sold in large quantities [44,45]. Herein Cyrene5 has been assessed for its suitability in substituting for NMP in the fabrication of PES membranes.

Materials.
The flakes of Ultrason5 E3020 P Polyarylethersulfone ( Figure 2(4)) of 55,000 Da and a powder of PVP Luvitek5 K-90 Pulver (Figure 2(5)) with 1,500,000 Da were obtained from INGE.BASF, Germany. The solvent Cyrene5 was supplied by Circa Sustainable Chemicals Ltd., UK. All the other solvents used in this study were of reagent grade and purchased from Merck Co., UK, and VWR Chemicals 5, UK. Deionised water (DI) was provided in-house by the lab using an ELGA CENTRA5 system. All chemicals were used without any further purification.

HSPiP's Predictive Power.
Hansen Solubility Parameters (HSP) [46] were chosen to predict solubility of PES in different solvents by mapping the three values (dispersion interactions D, dipolarity P, and hydrogen bonding ability H) in a three-dimensional ''Hansen space", using 5 th edition 5.0.03 of HSPiP (Hansen Solubility Parameters in Practice).
When determining whether a solvent will dissolve a sphere, it is useful to calculate the relative energy difference, RED: where Ro is the radius of interaction of a Hansen solubility parameter sphere determined experimentally ( Figure 1) and Ra is the solubility parameter distance between polymer and solvent and can be calculated from A 3D representation ( Figure 1) shows the distance of the solvents which dissolved the polymer and PES3020. A nonsolvent will have a Ra larger than the sphere's Ro, making  RED > 1, while solvents likely to affect the material will have a RED < 1. A solvent with a RED < 1 and closer to the polymer's centre (with D=20.42, P=9.89, and H=6.55) is predicted a better solvent. Cyrene5 has the smallest distance from polymer in the Hansen space, suggesting the greatest affinity to PES3020 (Table 1). A full list of the solvents used can be found in the Supplementary Material accompanying the manuscript (SM Table 1). As seen in Table 1, comparing HSP of Cyrene5 and the other polar aprotic solvents, it was found that Cyrene5 has a dispersion close to DMSO (18.9 vs. 18.4), a polarity parameter close to NMP (12.4 vs. 12.3), and a hydrogen bonding close to NMP (7.1 vs. 7.2). The similarity of Cyrene5 to the other polar aprotic solvents has been studied previously [40].

Preparation of Casting Solutions.
In this study, the polyethersulfone membranes (PES/C and PES/N) with highly asymmetrical pore structure were fabricated from hot (70 ∘ C) and cold (17 ∘ C) casting gels of PES, PVP, and solvent (Cyrene5 or NMP), using a nonsolvent phase inversion technique (NIPS). Table 2 shows the formulations of PES ultrafiltration flat sheet membranes using two different solvents and with/ without PVP.
In this study, some of the characteristics of Cyrene5 and NMP are presented in Table 3, showing the difference between the two solvents.

Membranes Fabrication.
The casting solution was prepared by dissolving 20 wt% of PES pellets (as compared to the mass of solvent) into Cyrene5 or NMP at a temperature of 70 ∘ C for 4h. Different concentrations of PVP were added under continuous stirring ( Table 2). The casting solution was degassed and then placed on an acrylic plate for PES/N or on a Polyester (PET) nonwoven fabric CraneMat CU632 (Neenah Technical Materials, USA) for PES/C. The casting gel was cast using a RK Print K101 bench casting machine (RK Print, UK) at a speed of 3 cm s −1 for the membranes produced in NMP and 2 cm s −1 for membranes produced in Cyrene5. The speed of the casting machine was found to have a significant influence over membrane filtration and a slower speed was considered for PES/C due to its higher viscosity. The thickness of all membranes was controlled at 500 m. The casting film was submerged in a coagulation bath containing deionised water at RT, causing the PES to precipitate. Membranes were then washed three times in distilled water for 10 minutes while under sonication in order to wash out residual solvent the fabricated membranes were then stored in deionised water until further use. The membranes were prepared at RT of 17 ∘ C and a humidity of 74-78%. The temperature of water bath was 12 ∘ C and the time from casting the gel to placing it in a water bath was limited to a maximum of 5 seconds.
Due to the higher density of Cyrene5 and the amount of PVP used, both of which add to the total viscosity of the system; the gel was cast when hot. The viscosity of PES/C10 was tested using a Brookfield R/S plus Rheometer and shows a dramatic drop during the heating to 110 ∘ C (SM Figure 1), where its behaviour tends towards a Newtonian fluid.

Membrane Characterisation.
To characterise the prepared membranes, they were first washed with deionised water and dried in vacuum oven at 80 ∘ C for 12 hours. The effects of the more viscous Cyrene5 on the pore size, porosity, surface morphology, and mechanical properties of PES/PVP membranes were studied by scanning electron microscopy (SEM), pure water permeability, ATR-FTIR, and thermal analyses. FTIR and TGA reflected the physical or chemical    changes on the surface or in the bulk of PES membranes after modification. ATR-FTIR spectra were recorded on a PerkinElmer Spectrum 400 FT-IR/FT-NIR Spectrometer with transmittance peaks in 4000-650 cm −1 region, with rapid scanning (4 scans) and resolution 4 cm −1 at room temperature.
Thermal stability of PES membranes was studied by TGA, giving weight loss of the produced membranes as a function of temperature. The membranes were heated from RT to 625 ∘ C at a rate of 10 ∘ C min −1 under a flow of nitrogen.
Scanning Electron Microscopy (SEM) images were taken on a JEOL JSM-6490LV, at 8kV from Bioscience Technology Facility, Biology Department, University of York, and were used to determine the morphology and structure of the membranes after they were frozen and fractured in liquid nitrogen followed by Au/Pd coating.
Porosity and pore size distribution test were determined using Mercury Intrusion Porosimetry (MIP) using a Micromeritics Autopore IV instrument located in the University of Leeds. This method is based on the behaviour of "nonwetting" liquids in capillary which cannot be absorbed by the pores of a solid itself, but requires an external pressure to be applied. By measuring the volume of mercury that intrudes into the sample material with each pressure change, the volume of pores in the corresponding size class can be obtained. MIP only shows accessible interconnected pores (if the closed pores are incompressible). The applied mercury pressure is inversely proportional to the size of the pores. A lower pressure is needed to penetrate large pores, while a greater pressure is needed to access smaller pores. From the pressure versus intrusion data, the instrument generates volume and size distributions using the Washburn equation: where D stands for pore diameter ( m), P is the applied pressure (psi), represents Hg-air surface tension (484 mN m −1 ), and is Hg-air-porous material contact angle (140 ∘ ). NMR spectroscopy was used as supporting evidence of the pore size distributions and was considered a more sustainable and less toxic method compared to the MIP method, which can distort the skeletal porous structure of a sample [47]. However, this method was used previously to characterise porosity, pore geometry, connectivity, and permeability of sandstones and carbonates [48,49]. In this project, the NMR was used to confirm the volume of fluid filling the pore space and a T2 distribution (equivalent to a pore size distribution, PSD) is obtained after deconvolution of magnetization relaxation ( Figure 3).
As a measure of membrane's hydrophilicity, the water contact angle (WCA) was measured via the sessile drop method [50] using a Theta Lite optical tensiometer at a room temperature of 22 ∘ C. A range of 0.5-1.5 L droplet sizes of water were placed on the membrane surface and the images were recorded using the automated OneAttension software. The static contact angles were measured at a minimum of three random locations and the mean values reported to minimise experimental error.

Pure Water Permeability Test.
Membrane permeability was determined by measuring the pure water fluxes using a stirred cell (Sterlitech HP4750). The filtration solutions were stirred magnetically at 300 rpm and a constant temperature of 25± 0.5 ∘ C. Rates of filtration were determined by continuously weighing the filtrate on an electronic balance connected to a data logger. A digital electronic balance from Ohaus (Scout Pro Range), with an accuracy of 0.01 g was used to continuously measure the weight of the permeate. Recording the weight at certain time intervals also allows the calculation of the permeate flux: where is the water flux (L m −2 h −1 (LMH)), m is the mass at a given time, is the density of water, is the area of the membrane (m 2 ), and t is the time (s). Figure 4 shows schematically the arrangement of the membrane filtration and flux measurement equipment. The filtration cell was pressurised using nitrogen gas (oxygen free) with pressure controlled by means of a regulator on the cylinder. The 50 mm diameter membranes were first subject to an initial pressure of 1 bar until a stable flux was evident, then the pressure increased to 5 and 10 bar.

Gravimetric Method of Water Adsorption.
This method is repeatedly used in literature to evaluate the porosity of the membranes [28,50] via determining take up of water. The membranes were vacuum dried at 80 ∘ C, before being weighed and their size and thickness measured. They were then immersed in distilled water at RT for 24h, before being weighed again, and the increase in mass registered. The %water retained in pores was calculated using the following equation: where Ww and Wd represent wet and dry masses (g) of the membranes, respectively, A is the membrane surface area (m 2 ), l is thickness of membrane (m), and is density of water at RT (998 Kg m −3 ).

Results and Discussion
3.1. Porosity and Pore Size Distribution. The membranes are coded based on the solvent used (PES/C is the membrane produced with Cyrene5 while PES/N represents the one with NMP) and the concentration of PVP used, e.g., "0" means a membrane produced with no PVP, while "0.1" through to ''10" means a membrane produced with 0.1% through to 10% PVP, respectively ( Table 2). For example, a PES membrane produced with Cyrene5 and 0.1% PVP will be referred to as "PES/C0.1" while the same membrane produced in NMP would be denoted "PES/N0.1". As seen in Figure 5, the Cyrene-based membranes presented higher porosity in comparison to NMP in all cases, with a maximum of 79% and 78.8% for PES/C0.1 and PES/C1, respectively. The membrane prepared with Cyrene developed porosity even when not using pore forming PVP (PES/CO with 76.7% total porosity), which demonstrates superior porosity than a NMP-based membrane using pore forming additive (PES/N0.1) with 76.3% porosity. The membrane with 10% PVP has the lowest total porosity (54.9%), and at this point the membrane exhibits a different morphology (phenomenon seen in SEM images). On the other hand, the highest porosity (76.3%), for the membranes produced with NMP is obtained with a PVP concentration of 0.1%. The lowest porosity for the membranes produced using NMP occurs for PES/N10 (44.6%).
It is also significant that 0.1% PVP is enough to produce good quality PES/C membranes, lowering the amount of sacrificial polymer represents an important starting point for the sustainable membrane fabrication together with substitution of a toxic petroleum-derived solvent with a safe, bio-based alternative.
Both Cyrene-and NMP-based membranes present a slightly decreasing porosity when 0.5% PVP is used (PES/ C0.5 and PES/N0.5) and clearly decrease when using a higher concentration of PVP (PES/C5 and PES/N5). This decrease in porosity is more visible in the case of NMP-based membranes. However, this means that 1% of additive is clearly sufficient to give the maximum porosity in both types of membranes, with Cyrene-based membranes more superior than the NMP equivalents. Figure 6 shows the pore diameter of membranes produced with Cyrene5 and NMP. PES/C membranes developed bigger pores than PES/N, except for PES/C10, which has a different morphology. Interestingly, the pore size of a Cyrenebased membrane with no additive (PES/C0) has a similar pore diameter to that obtained in NMP-based membrane using at 5% PVP (PES/N5). This means that no additive is necessary to form pores when using Cyrene5 as solvent. Similarly a membrane produced using Cyrene5 and 1% PVP (PES/C1) shows the same pore diameter as a NMP-based membrane when using 10% (PES/N10) of the same pore forming agent. That means that less sacrificial pore forming agent is required when using Cyrene5 as solvent compared to NMP-based systems.
Membranes prepared with Cyrene5 have pore diameters from 0.09 m (PES/C10) to 1 m (PES/C1 and PES/C5), while the membranes prepared with NMP range between 0.2 m (PES/N0, 0.1, 0.5, 1 m) to 1 m (PES/N10). Figure 6 clearly shows similarity between PES/C1 and PES/C5 to PES/N10 in pore size and pore distribution (SM Figure 2 and SM Figure  3). The membranes made using Cyrene5 need 1% PVP to develop the biggest pores (PES/C1), while the membranes produced with NMP need twice as much PVP to develop the same pore diameter.
A high degree of variation in the pore system is seen when PVP concentration is increased. This is observed by the greater decrease in total porosity and an increase of pore diameters in the range between 0.5 and 1 m for PES/C0-5, followed by a dramatic decrease of pore diameter for PES/C10. The Cyrene-based membranes pore diameters follow a trend from a minimum with PES/C0 to a maximum with PES/C5. Above a 5% PVP loading, this type of membranes drops drastically in pore diameter. This is as a result of the morphology of the membrane changing significantly, with large cavities containing very small pores on the walls (as clearly seen in Figure 10).
This morphology is clearly seen in PES/C10 where a 'diffusion membrane' has been formed, at which point it is believed that the pores of 0.09 m are incorporated in the walls of cavities, replacing the classical macrovoids. The asymmetric membranes have variable pore diameters, from 0.001 to 10 m while the symmetric one (PES/C10) present more uniform pores, with majority of pores of 1 m diameter. PES/C0 presents a small quantity of pores of 0.1 m; the quantity of this type of pores increases with increasing of PVP in casting solution and their presence is exclusively in PES/C10.
Membranes made using NMP increase their pore diameters with increasing PVP concentration from PES/N0-10. The NMP-based membranes present the same small pore diameter until a concentration of 5% of pore forming additive is added to the casting gel, leading to a bigger pore diameter with a maximum for PES/N10.
The results from NMR spectroscopy are consistent with the previously reported MICP results. The largest pores (around 10 m) can be seen in the membranes produced with Cyrene5, increasing in line with the concentration of PVP used, with a maximum in PES/C5. For PES/N the number of larger pores increases with increasing PVP concentration, while smaller pores are no longer seen (e.g., pores in the range 0.001 to 0.01 microns are only observed for PES/N0.1). After deconvolution of magnetization relaxation of Nuclear Magnetic Resonance (see Figure 7), the obtained T2 curves confirm the volume of fluid filling the pore space and indirectly the pore size distribution, described above by MICP method.

Membrane Functional Groups via FTIR Spectroscopy.
The membranes present functional groups specific to PES, S=O symmetric stretch at 1149 cm −1 , C-SO 2 -C asymmetric stretch at 1320 cm −1 , C-O asymmetric stretch at 1240 cm −1 , C 6 H 6 ring stretch at 1578 and 1486 cm −1 , C-H stretch at 3096-3094 cm −1 , in addition to the residual PVP as indicated by a C=O stretch at 1667-1662 cm −1 , pyrrolidinyl radical at 1462 and 1423 cm −1 , C-N vibration at 1072 cm −1 , and C-H asymmetric stretch at 2950 cm −1 .
During the membrane precipitation step, some of the PES from the surface leaves the membrane together with most of the water-soluble polymer, but enough PVP remains (with its presence at 1667 cm −1 in Figure 8), providing the polymer surface with a more hydrophilic nature than the surface of PES alone.
In the case of PES/Cyrene5, coloration due to release of PES into the gelation media was less apparent than in the PES/ NMP systems. This indicates a stronger PES/PVP interaction in Cyrene5 (SM Figure 4).

Scanning Electron Microscopy (SEM) Analysis.
All membranes produced herein (except for PES/C10) present a typical structure with a thin top layer supported on a spongelike substructure (important for the mechanical resistance of the membranes), and macrovoids, due to the instantaneous demixing during the phase inversion [51], where PVP behaves as an antisolvent agent in the demixing step due to its high solubility in water. The casting solution has a high affinity for water which can cause changes in morphology and performance [4]. In this study the gel was cast and the glass plate with the casting gel was quickly immersed in the coagulation bath meaning the exposure to the atmosphere was less than 5 seconds. Finger-like structures can be seen near the inner (connecting the top layer to the sponge-like structure) and outer surface, while sponge structure can be seen in the centre of membranes. It has been generally accepted that instantaneous demixing leads to macrovoid structure and delayed demixing leads to a sponge-like structure [51,52].
While the sponge-like structure looks the same in both Cyrene5 and NMP producing membranes, Figure 9 shows the difference between them in a macrovoid layer at the same magnification.
When pure PES is used, the phase separation of polymer solution occurs immediately, due to low viscosity of the casting gel. The higher viscosity of Cyrene5 (14.5 m Pa s at 20 ∘ C) compared to NMP (1.67 m Pa s at 20 ∘ C) adds to the casting gel's viscosity leading to a slower demixing process and hence to a more visible sponge-like structure ( Figure 10). The modification of polymer solution (by adding PVP) changes the pore structure of the membrane [53]. Also, different temperatures of the casting gel lead to differences in their morphology ( Figure 10): The membranes cast from the cold gels present more finger-like layers than the plain PES/C0 (which is more viscous), especially in the middle of the membrane. The membranes cast from Cyrene5 hot gels showed more spongelike structure, with macrovoids all the way through the membrane. A greater difference in morphology starts from a concentration of 5% PVP, with the greatest at a concentration of 10% PVP. PES/C10 from a hot gel (Figure 10(l)) developed a completely different surface morphology, a symmetricallike structure, looking like a porous web or sponge. This could be caused by the large amount of PVP which slowed precipitation. When cast from a cold gel, the membrane shows asymmetrical layers with bigger space (Figure 10(i)). Increasing the PVP content in the casting solution led to slower demixing, especially towards the bottom layer of membrane (closest to the glass slide). When the exchange from solvent to antisolvent occurs, it starts at the surface, where finger voids are observed and slows as it approaches the bottom layer, when more sponge-like structures are seen. This structure presents pores which increase in diameter from the top to the bottom surface of the membrane. The rheology of the casting gel of PES/C10 was studied, with a decrease in viscosity with increasing temperature clearly visible (SM Figure 1). This had a great impact on the membrane morphology and differences in the permeability data are expected. A clear difference between the membranes cast from hot/cold gels is seen in NMP's case too (red circle in SM Figure 5) where the one cast from a hot gel presents a middle sponge layer, while the one cast from a cold gel clearly shows the finger layer all the way through the membrane.

Thermal Stability of Membranes.
The results of thermogravimetric analysis on the membranes prepared here are shown in Figure 11 and Table 4. PVP has a decomposition temperature around 440 ∘ C, while PES3020 decomposes at 580-600 ∘ C, which means that PVP would lose more weight before PES began to decompose.
The first degradation took place below 120 ∘ C due to moisture loss, followed by Cyrene5 volatilisation at around 170 ∘ C and 140 ∘ C for NMP, followed by PVP decomposition at ca. 420 ∘ C and PES decomposition at ca. 580-600 ∘ C. The differential thermogravimetric data shows a greater amount of residual PVP from PES/C membranes.
PES/C membranes consistently show a higher thermal stability than PES/N equivalents. PES/C0 has a thermal decomposition of 588.7 ∘ C, while PES/N0 decomposes at 578.4 ∘ C, which means that a greater energy is required to break the bonds of PES/C membrane. The peak of the 1 st derivative (inflection point) indicates the point of greatest rate of change on the weight loss curve, as seen in Figures  11(b) and 11(d).
The addition of PVP to the membranes decreases the thermal stability, due to a lower decomposition temperature of PVP than of PES, but the miscibility of both polymers was confirmed from the thermal analysis. The differences between the concentration of PVP and TGA residue may be due to PVP remaining in the membranes after washing with water, depending on the membranes porosity and pore size. It was found that PES/N membranes lose some PES from the membrane surface when casting the film. The same membranes lost more PES and PVP than PES/C, which means that Cyrene5 makes a better media for this type of filtration as well as being more hydrophilic. Figure 12(a), the static water contact angle of the new PES films produced with Cyrene5 decreased from 73.1% for PES/C0 to 37.5% for PES/C10, with a hydrophilic character in all cases. This decreasing of the contact angle indicates increasing in hydrophilicity of these membranes with increasing residual PVP content.

Surface Wetting Property of the New Membranes Produced with Cyrene5. As shown in
The wetting changes were recorded over time by means of the same methodology used for static contact angle, but with values taken over a 300 second range. It can be seen in Figure 12  manner and reduce their contact angle. A larger initial drop occurs as water rapidly fills pores at the surface of the membrane, until locally saturated. This phenomenon was found to be due to a result of the advancing angle (higher angle which relies on the hydrophobic surface component) changing into a receding angle (low contact angle of the hydrophilic component) upon a decrease in droplet volume and is due to imbibition and evaporation. [54] The contact angle of PES/C0 recorded the smallest change during the test time, due to a less hydrophilic character and the tendency of the droplet to be repulsed by the membrane surface at least at the beginning. Over the time, when the membrane is locally saturated, the droplet moves into the porous membrane and the contact angle changes. On the other side, PES/C10 membrane was found to have the most dramatic change in its contact angle, due to the most hydrophilic character and its porosity. The SEM images of the PES/C10 demonstrated a very porous membrane which would allow the droplet of water to be completely absorbed in under 60 seconds during the WCA experiment.

Pure Water Permeability.
Pure water permeability (PWP) testing was evaluated based on water flux. Evidence in the literature indicates that the sponge layer filters are slower than the finger layer of a membrane [51].
Based on these criteria, the permeate flux seen in Figure 13 shows differences between the membranes depending on the solvent used, the concentration of PVP used, and temperature of the casting gel. The membranes produced using Cyrene5 and cast from a cold gel show an increasing permeability when PVP is added with a maximum at 10% PVP, but with no permeability when using 0.5, 1, and 5% PVP. This is explained by the morphology of the membrane, with more finger layers on top of the membrane a dense sponge layer at the bottom of the membrane. This is due to the thickness of membrane chosen in this project (500 m) which results in different morphologies most likely depending on speed of exchange between the solvent and water (SM Figure 6). At the top surface, this is rapid, with the polymer precipitating out and voids forming as a result of water displacement. At the bottom, crystallisation has dominated, with fewer voids. This results in low permeability as the finger regions are not interconnected. The concentration of PVP used had a great impact of permeability when using over 0.1% PVP, when the produced membranes showed no flux, due to the presence of dead ends of the finger layers presented in their morphologies. The new membranes PES/C0 and PES/C0.1 cast from cold gels together with PES/C0 from a hot gel are believed to be suitable for a ultrafiltration (UF), while a permeability is lower than 20 LMH/bar for a nanofiltration/reverse osmosis (NF/RO), while the PES/C10 membrane showed greater fluxes more in keeping of a microfiltration membrane (MF). Instead, when the same membranes are cast from a hot gel, the produced membranes showed a permeability with a maximum at 10% PVP, but with a smaller permeate flux than the corresponding cold gel membrane. The membranes produced from a cold gel, using Cyrene5 and a PVP concentration above 0.1%, developed a morphology with finger layers in the middle of the membrane but are not interconnected to the surface, thus registering no permeate flux. When adding PVP, the permeability also increased in PES/N, with a maximum flux at 10% PVP and when the membrane was cast from a cold gel. When cast from a hot gel, the membrane produced from the same concentration of PVP showed a higher flux of water due to a void in the middle of the membrane which permitted a higher permeability. Looking at Figures SM 5(b) and 5(e) relating to 0.1N, the hot cast seems to have a layer in which all the pores appear blocked about 1/2 way through the membrane (red circle in SM Figure 5(e)). However, in the cold cast, the pores appear to traverse the membrane (red circle in SM Figure 5(b)). The presence of the middle sponge layer in PES/N membranes which were hot cast showed a lower permeability than the ones coming from a cold gel, except for PES/N0.5, where the morphology shows a thicker sponge layer, which would generally slow the flux down. Figure 14 shows the absorption capacity of PES/C (in blue) and PES/N (in orange) to retain pure water in their pores. Although less precise than other methodologies applied in this research, this is the standard test to evaluate membrane porosity [17]. Figure 14 shows that PES/C consistently developed a greater water retention capacity than the respective PES/N membrane (consistent with MICP and SEM data). MICP and gravimetric data present differences between the samples generated using the same solvent. For example, gravimetrically PES/C5 presents the biggest capacity of water absorption, while for MICP data, PES/C0.1 has the biggest total   porosity. However, PES/C10 showed the smallest capacity to retain water, which is entirely consistent with MICP and SEM data.

Conclusions
Polyethersulfone flat sheet membranes for water filtration applications were prepared by the NIPS technique employing Cyrene5 as an alternative solvent and compared to PES produced using the traditional NMP. Membrane morphology and performance were tailored acting on the casting solution composition and temperature. The new membranes produced with Cyrene5 are more sustainable, with less of both polymers' loss and tunable pore size and contact angles, from a less hydrophilic PES/C0 to a more hydrophilic PES/C10 when the hydrophilic additive is added to the casting solution. The produced membranes with the biobased solvent Cyrene5 showed a greater total porosity, bigger pore size, and higher thermal stability, compared to the PES membranes produced with NMP. It was found that no additive is necessary to form pores when Cyrene5 is applied as solvent, PES/C0 having the same pore diameter as NMP-based membranes produced with 5% PVP (PES/N5), meaning that no additive many not be necessary for the role of pore forming when using Cyrene5. PES membranes produced with Cyrene5 and 0.1% PVP (PES/C0.1) and 1% PVP (PES/C1) developed the largest porosity (ca. 79%) for the both Cyrene5 and NMP-based membranes. PES produced with Cyrene5 and minimal concentration of PVP (0.1% PVP) developed the same pore diameter as PES produced with NMP at a higher concentration of PVP (5%) and the largest pores are registered in PES when using Cyrene5 with 1% PVP, while for a membrane produced with NMP, the largest pore diameter appears using a higher concentration of PVP (10%). The permeability of the new membranes produced with Cyrene5 was easily tailored in the range between NF/RO to MF, by changing the temperature of the casting gel, whereas only slightly small changes were observed when using NMP. While the PES/N showed a trend with a maximum filtration for PES/N cast from a cold gel when 5% PVP was used, the same membranes cast from a hot gel showed no supporting trend, probably because NMP evaporated quicker than Cyrene at elevated temperatures. Completely new morphologies were found in PES produced in Cyrene5, when using 10% PVP, regardless if cast hot or cold. When using Cyrene, the morphology was different for the two casting gels, with a symmetrical one observed when hot cast; this is less obvious when cast from a cold gel. Both hot and a cold cast PES/C10 have showed the best filtrate flux, with the potential, for application in microfiltration. Utilising cold casting, minimal addition of PVP leads to better filtration properties than the corresponding membrane produced without any additive; this suggests PES/C0.1 has applications as a hydrophilic, highly biocompatible/haemodialysis ultrafiltration membrane. The relatively high viscosity of Cyrene5 adds to the overall viscosity of the PES/PVP casting gel and this can be exploited to tailor membranes to different ranges of filtration with different physical properties. Decreasing viscosity with temperature is a potential factor to consider when casting membranes containing Cyrene5.

Data Availability
Data supporting this publication is available in the supporting information while additional raw data is available via https:// doi.org/10.15124/c6545fa3-c5aa-4741-9ad0-1fe092f292cf.

Conflicts of Interest
The authors declare that they have no conflicts of interest.

Supplementary Materials
Supplementary information is available and contains a table listing all the solvents studied via HSPiP (SM Table 1); a plot for the temperature dependent viscosity of the PES/C10 sample over the range of room temperature to 110 ∘ C (SM Figure 1); MICP pore diameter distributions for each sample prepared in Cyrene5 (SM Figure 2) or NMP (SM Figure  3); a photograph showing the coagulation bath of PES/PVP membranes in Cyrene5 and NMP (SM Figure 4); SME crosssection images of each membrane prepared in NMP (SM Figure 5); SEM cross-section images of PES/C1 at different thickness (SM Figure 6). (Supplementary Materials)